Method and device for retrieving biomechanical parameters of a stride
A method and device determine biomechanical parameters of a runner from an accelerometer worn off-torso on a wrist, upper arm, head, or shoe. Acceleration including a vertical component is acquired and cadence is determined from the initial sequence. At least one frequency component caused by relative motion of the device with respect to the runner's center of mass—comprising a fundamental at approximately one-half the cadence and one or more harmonics—is identified. A cadence-locked comb filter, whose notch center frequencies are functions of the cadence and that applies greater attenuation at the one-half-cadence fundamental and its harmonics than at the cadence fundamental, is applied to generate a modified acceleration sequence that approximates center-of-mass vertical acceleration. Biomechanical parameters such as contact time, flight time, stiffness, undulation, and take-off or landing angle are then determined from the modified sequence.
1 . A method for determining biomechanical parameters of a stride of a runner, the method comprising:
measuring an initial sequence of acceleration data in at least a vertical direction using an accelerometer device that is worn on a wrist, upper arm, head or shoe of the runner;
determining a cadence of the runner from the initial sequence of acceleration data;
identifying in the initial sequence of acceleration data at least one frequency component caused by relative motion of the accelerometer device relative to a center of mass of the runner, the at least one frequency component comprising a fundamental at approximately one-half the cadence and at least one harmonic thereof;
attenuating the frequency component, by applying a cadence-locked comb filter whose notch center frequencies are functions of the cadence and that applies greater attenuation at a one-half-cadence fundamental and its harmonics than at a cadence fundamental, so as to determine a modified sequence of acceleration data that approximates center-of-mass vertical acceleration; and
determining the biomechanical parameters from the modified sequence of acceleration data.
2 . The method of claim 1 , wherein the identifying comprises:
extracting the cadence directly from the initial sequence of acceleration data; and
using the cadence for determining the frequency component.
3 . The method of claim 1 , wherein the identifying further comprises:
extracting a sample window from the initial sequence of acceleration;
performing a FFT of the sample window to form a transformed signal; and
determining the most significant frequency peak from the transformed signal.
4 . The method of claim 2 , further comprising:
verifying that the cadence determined directly from the initial sequence of acceleration data corresponds to a frequency peak.
5 . The method of claim 2 , wherein the identifying further comprises retrieving at least one harmonic peak from the cadence and/or from the most significant peak,
and wherein the step of attenuating comprises attenuating the harmonic peak.
6 . The method of claim 2 , wherein the attenuating comprises attenuating frequencies at half the frequency of a cadence peak.
7 . The method of claim 2 , wherein the attenuating comprises attenuating frequencies at some harmonics of a cadence peak.
8 . The method of claim 2 , wherein the attenuating comprises attenuating a cadence peak by a first attenuation factor, and applying a greater attenuation to harmonics corresponding to an arm swing peak such that attenuation at the one-half-cadence fundamental and its harmonics is greater than attenuation at the cadence fundamental.
9 . The method of claim 5 , wherein frequencies between two attenuated harmonics are less attenuated than the harmonics.
10 . The method of claim 1 , further comprising:
extracting a sample window from the initial sequence of acceleration;
performing a FFT of the sample window, so as to compute a FFT signal; and
inputting the FFT signal to a machine learning system, such as for example a neural network, in order to determine the cadence of the runner.
11 . The method of claim 1 , wherein the measuring is performed in a wristwatch, while the steps of identifying, attenuating and determining are performed in a smartphone or in a computer.
12 . The method of claim 1 , wherein the biomechanical parameters include at least one among:
flight time, contact time, lowering, elevation, undulation, asymmetry, regularity, reactivity, stiffness, take-off angle, and/or landing angle.
13 . The method of claim 1 , further comprising:
using the at least one frequency component caused by relative motion of the accelerometer device relative to the center of mass of the runner, in order to determine an arm movement.
14 . The method of claim 1 , the initial acceleration data being measured in a plurality of orthogonal directions, the attenuation being performed independently for each direction.
15 . The method of claim 14 , wherein the attenuated frequencies are the same for each direction.
16 . A non-transitory computer product containing program code for causing a processor to execute the method of claim 1 when the code is executed.
17 . The method of claim 1 , further comprising validating the modified sequence of acceleration data by comparing a spectrum of the modified sequence to a stored center-of-mass spectral template and proceeding to determine the biomechanical parameters only when a similarity metric between the spectrum and the template exceeds a threshold.
18 . An accelerometer device for determining biomechanical parameters of a stride of a runner with a high accuracy, comprising:
an accelerometer that measures an initial sequence of acceleration data in at least a vertical direction;
a memory; and
one or more processors communicatively coupled to the accelerometer and the memory, wherein the one or more processors are collectively configured to:
identify in the initial sequence of acceleration data at least one frequency component caused by relative motion of the accelerometer device relative to center of mass of the runner, the at least one frequency component comprising a fundamental at approximately one-half the cadence and at least one harmonic thereof,
attenuate the frequency component, by applying a cadence-locked comb filter whose notch center frequencies are functions of the cadence and that applies greater attenuation at a one-half-cadence fundamental and its harmonics than at a cadence fundamental, so as to determine a modified sequence of acceleration data that approximates center-of-mass vertical acceleration; and
determine the biomechanical parameters from the modified sequence of acceleration data.
19 . The accelerometer device of claim 18 , further comprising:
a battery; and
a display.
20 . The accelerometer device of claim 18 , wherein the one or more processors are further collectively configured to adaptively update notch center frequencies and bandwidths of the cadence-locked comb filter in response to a detected cadence change exceeding a threshold percentage of a current cadence.